How can AI improve your R&D?

According to MIT, AI was a suitcase word which fit numerous concepts. An example of suitcase word is "strategic," so when people say, " I am strategic," many often scratch their heads questioning what does that even mean? To simplify the complexity, the definition of AI is how to use Descriptive Data to go from Predictive Analytics to a Prescribed Action. A conspicuous example is Tesla which uses lots of digital images as its descriptive data to predict whether or not this is a human, a pet or not to prescribe an action which can be either dodge or drive straight for its self-driving technology.

While I have not seen the exact use case, but summarizing the research papers using AI tools would be useful for any researcher in the preliminary stage of their investigation.

Although artificial intelligence has relatively recently entered the household, it has been in the R&D universe since the 18th century with "Mechanical Turk". If you are on an R&D team, you are most likely already benefiting from artificial intelligence in one way or another. However, recent rapid developments in the field of machine learning inevitably attract the attention of many different fields. Although the products on the stage are very exciting, the fact that machine learning works as a black box for fields that do not rely on probabilistic decision-making, such as medicine or law, still shows that it cannot penetrate such fields very much.

Tech Support here resides in R&D because when launching new products, R&D wants to know in real time if there's a customer complaint. Unlike the old days when multiple headcounts were loaded up to cover three shifts, especially during potential peaks hours when all hoped zero customer complaint, one could recognize the inefficiency. Today with regenerative AI and Natural Language Processes (NLP), chatbot AI can converse like a human being with natural voice in local languages. When global customers want to talk live operators, this queuing problems can be solved since all know average service rate in minutes, lamda for arrival rate, Poisson distribution & exponential distribution for Heijunka level loaded staffs including peak times.

It normally depends with the scope you are working on and its applicability.In general A.I will accelerate your processes in terms of time resource, cut down on human resource, meet deliverables & deadlines on time, quality of your research, efficiency, reliability, if well utilized it reduces error margins of your research, proving concepts and backing them up with valuable and reliable supporting data to be acted upon. As they say calculators is for Mathematicians so is AI for R&D.

In R&D process AI play a vital role all of the time,by identifying patterns, correlations, and emerging technologies, generative AI algorithms can assist R&D professionals in identifying potential areas of innovation and predicting future trends. This can help organizations prioritize their R&D efforts, allocate resources effectively, and stay ahead of the competition.Artificial intelligence technologies can help innovation teams overcome these challenges by analyzing large datasets, identifying patterns, and generating new ideas based on those configurations.

AI will revolutionize R&D as a force multiplier. The following tasks are only a partial list of what can be enhanced by AI at a fraction of the current cost: - Prior art search - Patent search and freedom to operate - Insights generation - Identifying white space for new products - Integrating with lab platform - Data analysis and experimental design - Predictive modeling - Generating insights and product ideas - Product design - Identifying market and product risks - Assessing competitive threats - Assessing regulatory risks - Estimating cost and price discovery - Reducing cycle time from product concept to implementation - Seamless integration with other business functions

Where products are based on formulas like food, cosmetics and many more, AI can help developing new formulas. For example the company Notco which produces vegan substitutes of products that uses animal based ingredients. Notcos AI System was able to help with new formulas e.g. to produces vegan milk - which contains of pineaple juice and cabbage. Interesting combination. Another solution is the creation of artificial avatars as test persons for user test like Google in an research project. They created a test environment for wireframes to check if an app prototypes use of screen elements is understood by the future users.

According to ASQ Six Sigma Black Belt materials, 100% inspection is only 85% effective because an inspector cannot detect a defect in a 8-hour shift with numerous similar parts. Five years ago, R&D leveraged three vision systems to do 100% inspection full day so that's nearly 300% inspection in the process. For robotic automation on XYZ axis, 2 digital libraries of attribute data (accepted & rejected parts) were built to apply the compare-and-comply technique. Over production periods, questionable parts were set aside by robots for a real human inspector to accept or reject units. That was how costly library amassed data. Last year, R&D used Google Tensor Flow to perform supervised learning with 100% inspection in real time. Voila! 🙌

The applications of AI in R&D are numerous; not only does AI help you to gather, organize and collate data from numerous sources, it also aids you in gleaning valuable insights from them. DeepMind's Alphafold is a sheer example of how AI can supercharge research process because it was able to produce outstanding results in protein folding based on the tons of data it was fed with. AI has even helped in discovering shortcuts to quantum algorithms, and will accomplish so much more if we only we harness its immense power!

Everyone out here will agree that use of AI can significantly reduce the time taken in R&D. However, though AI can be very helpful for general cases, it can not be relied upon in unique cases where novel methods are explored. In addition, the uniqueness which researchers bring in can not be guaranteed by AI. I also feel that excessive use of AI can give short term gains to humans, but they will impact the human intelligence in longer run. This can be even irreversible causing lower levels of innovation aspects from human brains. This can be eventually a huge risk for R&D. Nonetheless, a thoughtful adaptation of AI can enhance the R&D outputs.

Data management is vital in realizing the potential of AI and advanced analytics in R&D. A proper data set will capture more than what data is measured but also how those measurements were conducted. Having a full picture of the data and how it was collected is the first step in leveraging the data to its fullest potential.

AI has great potential in R & D. However, we need to watch out for challenges in terms of insufficient data sets and accuracy, hallucinations, and data privacy.

Based on my article in Harvard Business manager 2023-05: Data quality and availability: AI systems require large amounts of high-quality data to work effectively. It can be difficult to collect enough relevant and clean data. Data privacy and security: The use of cloud based AI-systems can be associated with data privacy and security risks. Biased Results: AI systems can reproduce prejudice and discrimination if they are trained on unevenly distributed data. AI Literacy: AI delivers not 100 percent perfect results. Once an output is delivered as human like sentence, nice image or good summary it will often taken for granted. Beeing paranoid as Andy Grove (IBM) put it, is a good attitude to keep quality high.

AI offers immense potential to elevate R&D processes, from data analysis to innovation. To start integrating AI, first clarify your R&D objectives, including the problems you aim to solve and expected outcomes. Second, evaluate your existing resources—data, infrastructure, and expertise—to identify gaps. Then, develop a project plan that is both structured and agile, and implement it. After deployment, closely monitor the performance metrics and adjust your AI models based on feedback and results. This ongoing iteration is key to leveraging AI effectively. Adopting such a methodical approach ensures not just the seamless incorporation of AI into your R&D but also maximizes its potential benefits.

AI can provide support in the front end of innovation, especially in the following areas: 1. analyzing real user data (sales, marketing, user ratings & comments, etc.) - to gain insight into actual customer needs | added value: time and quality gains; 2. Precisely tailored idea generation for corresponding user needs | added value: time savings and stimulation of own creativity, inspiration; 3. Structuring assistance in the development of concepts and business models - although AI does not produce magically great output - I can have the first steps automated and then generate a really good result through my technical expertise | added value: time savings. It is important to note, that not only GPT is aplicable here, but also ML-models.

Way back in the day, I successfully used what now would be called AI and data from fiber optic and other sensors to identify possible failure points on a simulated bridge element. AI then (based on artificial neural networks) and now is the same. The difference is that we now have incredibly capable computers and extremely large data sets. Many of us then considered AI to merely be a new form of curve fitting, excellent at interpolation within the range of the training set but totally unreliable outside of it. However, it might have at least one good use in carrying out R&D. If you ask it questions about the domain outside its training set, you cannot trust it, BUT, it might provide some new options to consider, out of the box, as it were.